Table 3: Second stage estimation results.

= + + + +
Comparison 1: nonmigrant sending against temporary migrant sending households
Number of Obs. = 212 HHs

Adj.
-stat.Const.
Coeff.Coeff.Coeff.Coeff.

0.842 (3,208) = 275.7***0.26***
(0.09)
2.971.13***
(0.072)
15.770.23***
(0.058)
4.08−1.32 
(0.835)
−1.59

= + + + +
Comparison 2: nonmigrant sending against permanent migrant sending households
No. Obs. = 177 HHs

Adj.
-stat.Const.
Coeff.Coeff.Coeff.Coeff.

0.911 (3,173) = 279.4***0.14*
(0.080)
1.721.68***
(0.075)
22.390.11***
(0.039)
2.79−4.35***
(0.666)
−6.53

refers to significant at 10% (); **refers to significant at 5% () and ***refers to significance at 1% ().
Bootstrap standard errors (as noted by Petrin and Train [17] and Karaca-Mandic and Train [18] bootstrapping methods applied for the entire two-step estimators provide a valid estimator of the covariance matrix which is similar to the estimation done to correct the asymptotic standard errors by programming the asymptotic formula of covariance estimates. Bootstrapping is a convenient way of obtaining the covariance matrix estimators with two-step estimators and it also provides better parameter estimates particularly for conditions when asymptotic sampling distribution is too difficult to drive in multistage estimations [19, 20]. In this research, bootstrapping method is applied for the entire procedure of the two-step estimators with 1000 replications using STATA software package) are in parenthesis, source: author’s estimation.